Beruflich Dokumente
Kultur Dokumente
1. GENERAL ASPECTS
As it is well-known, the occurrence of major seismic phenomena is a rare event
from a statistical point of view. Due to the very large time horizon that can be
taken into observation as against to registering events in artificial systems, as well
as the non-periodicity of these events, there is the possibility of interpretation and
statistical modeling of these seismic phenomena. In Romanian: Dragomir (2009),
Lungu (1999), Lungu and Arion (2000), Radulescu (2004).
The statistical studies regarding the earthquakes usually start from the fact that rare
events are best described using the exponential law if considering the succession
of time intervals between events, or Poissons law if it is intended to model the
frequency of earthquakes (Scuiu and Zorilescu, 1978; Johnson, Kotz and
Balakrishnan, 1994; Evans, Hasting and Peacock, 2000).
The easiness of using these two distribution laws, distinct in nature, consists of the
fact that they are defined by the same parameter, characterizing the same
phenomenon the behavior of a system in time, from both continuous and discrete
points of view. A previous study made on seismic phenomena in Romania (Voda
and Isaic-Maniu, 1983) covering the time period 1400-2000, has failed to confirm
the hypothesis of an exponential behavior, the confirmed model being the biparametric Weibull model.
1 =
(s )
2
3/2
4
moment of rank 2), and respectively 2 - Kurtosis coefficient: 2 =
(where
2
(s )
2
Although it is considered that seismic events are rare events from a statistical
point of view, thus with reduced probabilities of occurrence, this hypothesis is not
confirmed for Romania (Figure 1).
Figure 1 - The Distribution of Earthquakes,
betw een 1100-2010 in Rom ania
7
5
5
4
4
1
years
The minimum value in a 50 year interval was 0 (1350-1400), and the maximum
number of earthquakes 8 - was registered between 1850 and 1900. The value of
the second quartile was 2, and Q1 = 1 and Q3 =5 respectively.
x+
x
f (x ) =
x
1 +
(1)
F ( x ) = 1 +
x
(2)
In the case that = 0 the probability density of this bi-parametric model is:
/ )( x / ) 1
(
f (x : , ) =
[1 + (x / ) ]2
(3)
( x / ) = x
F ( x : , ) =
=
+ x
1+ ( x / )
1+ ( x / )
1
(4)
The kth raw moment exists only when k < , when it is given by
k /
sin (k / )
( )
E X k = k B(1 k / , 1 + k / ) = k
(5)
where B() is the beta function. Expression for the mean, variance, Skewness and
Kurtosis can be derived from this. Writing b = / for convenience, the mean is
E ( X ) = b / sin b, > 1,
(6)
(7)
Explicit expressions for the Skewness and Kurtosis are lengthy. As tends to
infinity the mean tends to , the variance and Skewness tend to zero and the excess
Kurtosis tends to 6/5 (see also related distributions below).
The quantile function is:
1/
p
F ( p; , ) =
1 p
1
(8)
The log-logistic has been used as a simple model of the distribution of wealth or
income in economics, where Gini coeffcient is 1 / (Kleiber and Kotz, 2003) it is
known as the Fisk distribution. The log-logistic distribution provides one
parametric model for survival analysis. The survival function is
S ( t ) = 1 F ( t ) = 1 + ( x / )
(9)
( / )( x / )
h (t ) =
=
S (t )
1+ ( x / )
f (t )
(10)
F ( x; , , ) =
(11)
(x )
1 + 1 +
1 /
(1 / +1)
(x )
1 +
(12)
F ( x; , , ) =
2
( x ) 1 /
1 + 1 +
again, for 1 + ( x ) / 0.
The shape parameter is often restricted to lie in [-1,1], when the probability
density functions is bounded. When > 1, it has an asymptote at x = / .
Reversing the sign of reflects the pdf and the cdf about x = 0.
See also:
when = / , the shifted log-logistic reduces to the log-logistic distribution.
when 0, the shifted log-logistic reduces to the logistic distribution.
The shifted log-logistic with shape parameter = 1 is the same as the generalized
Pareto distribution with shift parameter = 1 .
1 n
F ( x ) = I X i x
n i =1
(13)
Dn = sup F ( x ) F ( x )
(14)
S =
i =1
( 2i 1) ln F
N
( X i ) + ln (1 F ( X n+1i ) )
(16)
( )
Ei
i =1
m
2 = S
(17)
where Oi is the observed frequency for bin i and Ei is the expected frequency for
bin i and is calculated by
Ei = N (F (Yu ) F (Yi ))
(18)
where F is the cumulative distribution function and Yu and Yi are the upper and
lower limits for class i.
The test statistic follows, approximately, a chi-square distribution with (k - c)
degrees of freedom where k is number of non-empty cells and c - the number of
estimated parameters for the distribution +1.
Therefore, the hypothesis that data are from a population with the specified
distribution is rejected if
2 > 2 ,k c
where , k c is the chi-square percent point function with k - c degrees of freedom
2
The three applied tests (Kolmogorov-Smirnov, Anderson-Darling and PearsonFisher) confirm with a high confidence degree the log-logistic distribution, by
parameters:
= 1.6112
= 1.9994
=0
The Probability Density Function (pdf) for the estimated values of the parameters
is presented in Figure 2, the Cumulative Distribution Function (cdf) in Figure 3,
and hazard function in Figure 4.
Table 2 presents the values of the main indicators of the log-logistic distribution for
a number of x = 0, , 10 events.
Statistic
Functions
pdfprobability
density
function
CDF cumulative
distribution
function
h(x) hazard
function
S(x) distribution
10
0.3033
0.2077
0.1238
0.0757
0.0488
0.0333
0.0234
0.0171
0.0129
0.01000
0.2467
0.3193
0.5001
0.6579
0.7535
0.8545
0.8828
0.9033
0.9186
0.9304
0.3975
0.4116
0.4029
0.3533
0.3035
0.2295
0.2032
0.1819
0.1645
0.1499
0.7533
0.4998
0.3421
0.2465
0.1859
0.1455
0.1172
0.0967
0.8814
0.0696
Figure 2
Probability Density Function
0,32
0,28
0,24
f(x)
0,2
0,16
0,12
0,08
0,04
0
0
0,8
1,6
2,4
3,2
4,8
x
Log-Logistic (1,6112; 1,9994)
5,6
6,4
7,2
F(x)
0,56
0,48
0,4
0,32
0,24
0,16
0,08
0
0
0,8
1,6
2,4
3,2
4,8
5,6
6,4
7,2
5,6
6,4
7,2
x
Log-Logistic (1,6112; 1,9994)
Figure 4
Hazard Function
0,4
0,36
0,32
h(x)
0,28
0,24
0,2
0,16
0,12
0,08
0,04
0
0
0,8
1,6
2,4
3,2
4,8
x
Log-Logistic (1,6112; 1,9994)